Intel's tick-tock isn't coming back (www.theverge.com)

On Intel’s Q3 2025 earnings call, CEO Lip-Bu Tan and CFO David Zinsner made clear the company is prioritizing AI server chips over consumer processors as it manages chip shortages that should peak in Q1 next year. Intel posted a return to profit aided by lifelines from Nvidia, SoftBank and the U.S. government, but warned supply is tight and it will steer capacity, pricing and product mix to favor DCAI (data center AI) shipments while CCG (consumer) volumes will be modestly down. Intel also committed to releasing new AI GPUs annually — following Nvidia and AMD — and will ship only one Panther Lake SKU this year while pushing Lunar Lake parts in the near term. Technically, Intel says its 18A node is “adequate” for supply but not for target margins; yields need improvement and may not reach acceptable levels until 2026–2027. 18A will be a long-lived node for at least three generations, and Intel’s 14A program looks stronger in early performance and yield metrics. For the AI/ML community this means more short-term capacity for training and inference in data centers (good for scaling workloads), but constrained consumer GPU/PC availability and slower cadence in high-end client product rollouts. The shift away from “tick‑tock” toward node longevity and demand-driven capacity investment signals a more conservative, data-center-first Intel positioned to meet immediate AI demand rather than revive rapid consumer-generation cycles.